Darobactin B Stabilises a Lateral‐Closed Conformation of the BAM Complex in E. coli Cells

Abstract The β‐barrel assembly machinery (BAM complex) is essential for outer membrane protein (OMP) folding in Gram‐negative bacteria, and represents a promising antimicrobial target. Several conformational states of BAM have been reported, but all have been obtained under conditions which lack the unique features and complexity of the outer membrane (OM). Here, we use Pulsed Electron‐Electron Double Resonance (PELDOR, or DEER) spectroscopy distance measurements to interrogate the conformational ensemble of the BAM complex in E. coli cells. We show that BAM adopts a broad ensemble of conformations in the OM, while in the presence of the antibiotic darobactin B (DAR‐B), BAM′s conformational equilibrium shifts to a restricted ensemble consistent with the lateral closed state. Our in‐cell PELDOR findings are supported by new cryoEM structures of BAM in the presence and absence of DAR‐B. This work demonstrates the utility of PELDOR to map conformational changes in BAM within its native cellular environment.


Molecular biology
Full-length genes for BamABCDE in a pTrc99a vector, with a His 8 -tag on the BamE C-terminus (pJH114) [1] , was obtained from Harris Bernstein, NIH.To produce the expression plasmid used for in cell EPR work, this plasmid was modified to remove the His8 tag from BamE, and the natural BamA cysteines were mutated to serine (BamA C690S C700S), creating Cys-free BAM.Note that BamBCDE contain only a single N-terminal Cys that is acylated and used for membrane association of these lipoproteins.Cysteines were then introduced at the desired positions in BamA.All modifications were performed using the Q5 site-directed mutagenesis kit (NEB) according to the manufacturer's instructions.

Expression of BAM complex variants for in cell EPR
All growth used carbenicillin as a selection marker.Expression plasmids for cysteine-variants of the BAM complex with the natural cysteines from BamA mutated to serine (BamA C690S C700S) were transformed into BL21(DE3) cells and plated on agar.Single colonies were then inoculated into LB and grown overnight (37°C, 200 rpm).For expression, overnight cultures were diluted 1/100 into 100 mL of 2x TY broth in baffled flasks and allowed to grow (37°C, 200 rpm) until OD600 reached 0.5-0.6.BAM expression was then induced with 0.4 mM IPTG and allowed to proceed for 1.5 hours, before harvesting cells by centrifugation (5,000g, 10 min, Beckman JA-25.50 rotor).Supernatant was discarded and cells resuspended in pre-cooled spin-labelling buffer (50 mM MOPS pH 7.5, 50 mM NaCl) supplemented with 1% (w/v) D-glucose.Resuspension volume was 3 mL buffer for every 50 mL of the original cell suspension pre-harvest.Cells were kept on ice until used for MTSSL labelling (see In cell labelling of BAM dual-cysteine variants with MTSSL).

Western blots of cell suspensions expressing BAM for in cell EPR
For western blots of whole cells, the equivalent of 1 mL of cell suspension at OD600 1 was spun down (3,000g, 3 min) and the supernatant discarded.Cell suspensions were produced as described above, except for uninduced samples, where IPTG was not added.Pellets were either processed immediately or stored at -20 °C until needed.The pellet was resuspended in 100 μL of 1x loading dye (50 mM Tris-HCl pH 6.8, 1.5% (w/v) SDS, 0.001% (w/v) bromophenol blue, 10 % (v/v) glycerol) and boiled for 10 minutes (100 °C).The suspension was then spun (16,000g, 10 min) and 10 μL of the supernatant was loaded onto Mini-PROTEAN® TGXTM Precast Gels (Bio-Rad) and run at a constant voltage of 200 V. Gels were transferred onto PVDF membranes using a Trans-Blot Turbo transfer system (Bio-Rad), according to the manufacturer's instructions using the 1 mini-TGX transfer protocol.Membranes were then blocked overnight in 50 mM Tris-HCl pH 7.2, 150 mM NaCl, 0.1% (v/v) Tween-20 (TBST), 5% (w/v) milk powder on a roller at 4 °C before addition of primary antibody.For detection of BamA, rabbit polyclonal antibodies raised against its C-terminal peptide (CQPFKKYDGDKAEQFQFNIGKT) [2,3] (named αBamAC-term) was used at a 1:3000x dilution in TBST, 5% (w/v) milk powder and allowed to bind for 2 hours on a roller at room temperature.This polyclonal antibody was a kind gift from Harris Bernstein (NIH, USA).The membrane was then washed 3x with TBST, 5% (w/v) milk powder (5 min contact time per wash) before addition of anti-rabbit IgG, HRP-linked antibody (Abcam, UK) at a 1:10,000 dilution in TBST, 5% (w/v) milk powder.This was again allowed to bind for two hours and was then washed off with two washes of TBST and one of TBS (5 min contact time).The membrane was developed using SuperSignal TM West Pico Chemiluminescent Substrate (ThermoFisher).

Heterologous production of darobactin B
For heterologous production of darobactin B, the darobactin A core peptide codons of plasmid pZW-ADC3.2 [4]carrying two propeptide sequences were exchanged to WNWTKRF by PCR and subsequent recircularization in two rounds.PCR was performed using Q5 polymerase (NEB Biolabs) according to the manufacturer's instructions with the primer pair 5'-TGGAACTGGACCAAACGTTTCTAAGCATAATGCTTAAGTCGAACAG-3'/5'-GAAACGTTTGGTCCAGTTCCAAGCGGTGATTTCCGGGATTT-3'.The resulting fragment was gel purified on 1% (w/v) TAE agarose gels, recovered from the gel and recircularised using home-made isothermal assembly mix [5] Then, E. coli Top10 was transformed with the reaction using standard electroporation methodology and selected on LB supplemented with kanamycin (LBKan, final concentration 50 µg/mL).The assembled plasmid was isolated using standard miniprep methods and the process was repeated using the primer pair 5'-TGGAACTGGACGAAACGTTTCTAAAAGGAGAGCCATAAATGAATGC-3'/5'-GAAACGTTTCGTCCAGTTCCAGGCCGTGATCTCAGGGATCT-3'.Successful codon exchange was corroborated by sequencing and the final plasmid pNB-DarB3.2was introduced to DAR resistant E. coli Bap1.E. coli Bap1 + pNB-DarB3.2was cultivated in LBKan at 30°C and 200 rpm and darobactin B production was induced by addition of 1 mM IPTG.Darobactin B was purified from the medium according to Böhringer et al [6] .In brief, cells were separated from the medium by centrifugation after 2 days and the supernatant was extracted by addition of XAD16N resin.The resin was washed with H2O and eluted using 50% (v/v) MeOH/H2O + 0.1% (v/v) formic acid (FA) and the MeOH was evaporated using a rotary evaporator.The aqueous crude extract was subsequently separated using ion-exchange chromatography (SP Sepharose XL), eluting with 50 mM NH4Ac buffer in steps with pH 5, 7, 9 and 11.
Darobactin Bcontaining fraction was evaporated to dryness and final purification was performed on HPLC using a gradient of MeCN/H2O + 0.1% (v/v) FA.

In cell labelling of BAM dual-cysteine variants with MTSSL
E. coli cells expressing cysteine variants of the BAM complex were prepared according to Section Expression of BAM complex variants for in cell EPR.For labelling, bacterial suspensions were diluted in Eppendorf tubes to a total volume of 1 mL and an OD600 reading of 10, using ice-cold spin label buffer (50 mM MOPS pH 7.5, 50 mM NaCl).In addition to the cysteine variants, each batch of samples measured included cells expressing BAM with the natural cysteines mutated out, or relevant single Cys variants, to allow measurement of background labelling.As MTSSL is rapidly reduced by the cell suspension, with a half-life of minutes [7] , timings for each labelling step were kept as similar as possible between samples to make observed signals comparable for each batch.
To each Eppendorf 1 μL of 100 mM (1-oxyl-2,2,5,5-tetramethyl-3-pyrroline-3methyl)methanethiosulfonate (MTSSL) stock solution dissolved in DMSO was added, for a final concentration of 100 μM.Samples were placed on a thermal shaker at 25 °C and labelling allowed to proceed for 5 minutes.Cells were then pelleted (5,000g, 3 min, 4°C).Supernatant was discarded and the pellet resuspended in 1 mL ice-cold spin label buffer, before centrifuging again as before.For CW-EPR, the pellet was then resuspended in 20-50 μL ice-cold spin label buffer (this was kept constant for each batch of samples).A 100 μL Hamilton was then used to draw the resuspended cells into a PTFE sample loading loop (Hamilton) and load it into a 3 mm (OD) quartz tube (Wilmad® quartz (CFQ) EPR tubes, Merck) from which spectra were recorded (usually starting 17 minutes post-label).
For PELDOR measurements with and without DAR-B, pellet was resuspended in 45 μL spin label buffer.43.9 μL of this was then mixed with 1. Analysis of different samples with data acquired on different days shows high reproducibility of the results (once accounting for slightly different expression levels of BAM on in the different cultures (5-10 %) (Figure S8).

Continuous-Wave EPR (cwEPR)
Spectra were recorded at room temperature on a Magnettech MS-5000 (Freiburg Instruments) operating at X-band (9.2 -9.6 GHz) using a microwave bridge power of 10 mW and 10 dB attenuation.
A magnetic field sweep between 330 and 345 mT was used, with a sweep time of 60s, field modulation frequency of 100 kHz and field modulation amplitude of 0.1 mT.Spectra were averaged over 10 scans.

PELDOR distance measurements
All the pulsed and DEER/PELDOR experiments were carried out at 50K using a BRUKER ELEXSYS E580 pulsed spectrometer operating at Q-band and equipped with a Cryogen-Free Variable Temperature Cryostat (CF-VTC) from Cryogenic Ltd.The instrument is equipped with both a BRUKER 400U second microwave source unit and a SpinJet arbitrary waveform generator (AWG).The measurements were performed with a 150 W TWT Q-band amplifier and EN 5106QT-2w cylindrical resonator with typically total of 60 μL sample volume, as previously described [8,9] .Briefly, the resonator was systematically overcoupled during all the pulsed experiments and the Q-factor was maintained approximately the same for all samples.
Echo detected field sweep (ED-FS) measurements were carried with p/2-t -p pulse echo detection sequence with pulse lengths set at 16 and 32 ns respectively, an inter-pulse delay of t = 380 ns and a shot repetition time (SRT) of 3ms.
The DEER experiments were carried out using the standard dead-time free four-pulse sequence: p/2(observer)-t1 -p(observer)-t--p(pump)-t1+t2-t--p(observer)-t2-echo [10][11][12] where all are rectangular pulses.The echo intensity was monitored as a function of t.The pump pulse was applied either using a second microwave source unit or AWG, for all experiments, at the maximum of the ED-FS spectrum whereas the observer frequency was set at 80 MHz offset from the pump frequency.The observer p/2 and p pulse lengths were set at 16 and 32 ns respectively.The p pump was set at 12 ns for maximizing the modulation depth and keeping both the observer and pump excitation bandwidths well separated.The interpulse t1 set to 380 ns whereas t2 was adjusted depending on the strength of the refocused echo signal and on the distances expected.The SRT was set to 3ms for all DEER experiments and the averages were typically between 6 and 46 hours until sufficient signalto-noise was given.When using the second ELDOR source only a 2-Step phase cycling was used to remove the receiver offsets, whereas in the case of the AWG 8 step was used to remove echo crossing artifacts [13] .The interpulse τ1 was stepped eight times starting for its initial value 380 ns by 16ns and the corresponding time traces were added together to average out the deuteron modulations, which in some cases could be quite prominent in the time traces.These are typical experimental values commonly used for τ averaging to remove deuterium modulation effects during the data acquisition.

PELDOR data analysis
Distance distributions were determined from the dipolar time evolution data (experimental data) using two different MATLAB-based software programs.It was necessary to use a combination of analysis approaches given the low modulation depths, obtained in this work when using an experimental background function or a background dimensionality that deviates from the commonly used homogeneous three dimensional background (Table S1 and Fig. S6).The background correction is a major problem often met when analysing PELDOR data obtained for membrane proteins or systems which exhibit extensive aggregation phenomena due to the non-homogenous environment of the spins under study [14] .Measurements on Cys-free BAM samples had shown that the background contribution from intermolecular spins resulting from unspecific labelling of any other exposed extracellular sites of other OMPs present, fits quite well with a dimensional distribution below three (d<3).Analysis of the potential distances between BAM oligomers in the OM, such as potential BAM islands (mediated by BamB-BamB interactions [15] resulted in distances between 120 -160 Å (with BAM in the lateral open or closed conformations), ruling out any contribution to the observed EPR data.Docking two BAM molecules together (again with BAM in lateral or closed conformations) in all possible conformations also resulted in a low probability of complexes with distances less than 50 Å for the 501-501, 755-755 and 501-755 spin pairs (with only 0.7 % < 40 Å and 2.4 % < 50 Å).Hence non-specific oligomerisation is improbable and unlikely to affect the data analysed here.
In the present work, for the PELDOR data analysis we used two approaches, Tikhonov regularization method implemented in DeerAnalysis 2019 [16] and DEERNet [17] .We used specific D values determined by fitting the background dimension when using Tikhonov regularization and the quality of the fit was assessed based on the L-curve method and the shape of the Pake pattern whereas the second method, DEERNet, doesn't require any assumption on the background dimensionality.For both methods, the distance distributions were determined with their associated uncertainties corresponding to 95% confidence intervals (2s) obtained either by the DeerAnalysis built in validation tool or the implemented option in DEERNet that gives a measure of uncertainty of the resulting distance distributions represented by 95% confidence bounds.For the L501R1-S755R1 pair in the presence and absence of DAR-B, additionally we validate the associated results derived from these two methods by using the time trace of two control samples measured under identical conditions.These control samples used in Prior to the DEER data analysis, data points at the end of the time trace were cut off to remove any "2+1" end artefacts [18] appearing at the end of the PELDOR time trace.

Background correction and its impact on the modulation depth
The PELDOR data analysis for L501R1-S755R1:DAR-B and L501R1-S755R1 pairs has shown that the modulation depth is strongly dependent on the type of the background correction function (Fig. 2c,d and Figure S6).This has been previously reported [14] .Both experimental and homogeneous background correction functions (d = 2.3) resulted in similar distance distributions (Fig. S5c, S6 and S7).However, the modulation depths are quite different for both mutants, 2.8% and 1.4% for Cys-free background and 5.8% and 5.9% for homogeneous background (d = 2.3), respectively.The BAM Cysfree backgrounds yielded more reliable and well-defined Pake patterns (see the associated Pake pattern in Fig. S7), however this has impacted severely the modulation depth and resulted in a poorer signal-to-noise ratio (SNR).This confirms that the use of a Cys-free and/or single Cys mutant background correction functions are desirable especially in cases where the spin labelling efficiency is not a major issue (Figure S6).Additionally, we have tested a homogeneous three-dimensional background correction function (d=3) and the associated background corrected data are shown in Fig. S6.Although it yielded the largest modulation depth (11.6% and 7.5% for L501R1-S755R1-DAR-B and L501R1-S755R1 pairs respectively (Table S1), it resulted in multimodal distance distributions with a persistent second distance around 5nm for both mutants (Fig. S7a3 and S7b3).Their associated Pake patterns are not well defined, and their shape indicate that a part of the background is attributed to the biradical contribution [16] .

Expression and purification of wild-type BAM complex for cryoEM
Wild type BAM (BAM-WT) was expressed and purified essentially as described previously [22] .Briefly, E. coli BL21(DE3) was transformed with plasmid pJH114 (expressing Bam ABCDE, with a His8 tag on the C-terminus of BamE (BamABCDE-CT-8xHis) in a pTrc99a vector, kindly provided by Harris Bernstein [1] .
Grids were blotted for 6 sec at 4 °C and >90% humidity and plunge-frozen in liquid ethane with a Vitrobot Mark IV 480 (ThermoFisher).

CryoEM Imaging
Datasets were collected on a 300 keV Titan Krios electron microscope (ThermoFisher) in the Astbury Biostructure Laboratory operated with a Falcon4 detector in counting mode.For the BAM-WT dataset, were collected using a Selectris energy filter operating with a 10e -V slit at a nominal magnification of 165,000 yielding a pixel size of 0.74.A total of 6063 movies were collected at a nominal defocus range of -1.5 to -3.0µm and a dose rate of ~5.8 e-/pixel/s.For the BAM:DAR-B dataset (BAM-WT plus DAR-B), a total of 3732 movies were collected at a nominal magnification of 96,000x yielding a pixel size of 0.83Å with a nominal defocus range of -0.9 to -3.0µm and a dose rate of ~6.4 e-/pixel/s.

Supplementary Tables
1 μL DAR-B ([initial] = 12 mM) and 10 μL glycerol-d8 (Sigma-Aldrich) for a total volume of 55 μL with a final DAR-B concentration of 240 μM.50 μL was then loaded into EPR tubes and frozen as above.For the minus DAR-B measurement, the 1.1 μL DAR-B was replaced with DMSO (solvent vehicle for the DAR-B).

Figure 2
Figure 2 for the experimentally-derived background correction were Cys-free BAM in the presence of DAR-B dissolved in DMSO (for L501R1-S755R1 in the presence of DAR-B) and Cys-free BAM in the absence of DAR-B and presence of the same final volume of DMSO used in previous condition (for L501R1-S755R1 in the absence of DAR-B).Following the experimental background model correction we performed Tikhonov regularization.Prior to the DEER data analysis, data points at the end of the conformation.High-res particles were taken for further downstream processing.d) Final map used for model building, coloured by chain.Blue = BamA, green = BamB, yellow = BamC, orange = BamD, magenta = BamE, red = darobactin B. e) FSC plot, calculated in Relion 4.0, used to estimate global resolution.f) Final reconstruction filtered and coloured by local resolution, calculated in Relion 4.0.

Table 1 | Acquisition parameters for PELDOR experiments.
All pulse lengths were kept the same.2-step phase cycling was used when using the second microwave source and 8-step phase cycling when using the AWG.*Modulation depth values derived from processing the primary PELDOR data with a background with dimension distribution determined by DEERnet (d = 2.3).**Modulation depth values derived from processing the primary PELDOR data with a homogeneous three-dimensional background correction.***Spin labelling efficiency estimation and error estimation based on our experimental setup, where the pump pulse lengths are ranging between 12-16ns, the maximum modulation depths (lmax) range between 25-30% for 100% spin labeling efficiency of a biradical nitroxide sample.